yt-pipeline-n8n

# yt-pipeline-n8n

AI Summary

Purpose:

• Repo-specific memory for yt-pipeline-n8n, a personal YouTube Shorts

automation platform MVP.

Key points:

• Local path is ~/hw/project/yt-pipeline-n8n.
• README describes a Mock MVP for operating multiple Japan-target YouTube Shorts

channels without external APIs.

• Core surfaces are FastAPI backend, PostgreSQL, Redis, n8n, Streamlit

dashboard, Python pull worker, local storage, mock docs/upload, TTS/image providers, FFmpeg/Remotion rendering, and AI quality packs.

• The project has strong manual runbook evidence but no obvious automated test

files in the current scan.

Relevant when:

• Working on YouTube Shorts automation, control/worker plane design, prompt

quality packs, TTS/image/rendering, or onboarding material.

Do not read full document unless:

• You need environment keys, runbook details, or worker/provider routing.

Linked documents:

• ai/workspace/repos.md
• ai/wiki/projects/youtube-shorts-automation-platform.md
• ai/worklog/index.md

Repo Info

• Repo ID: yt-pipeline-n8n
• Main branch: main
• Local path: ~/hw/project/yt-pipeline-n8n
• Remote: [email protected]:hyunwook711/yt-pipeline-n8n.git
• Latest observed commit: 9973eb5 2026-05-09 Add AIGS longform worker scaffold

Common Commands

docker compose up -d
docker compose --profile worker up -d --build worker
docker compose logs -f backend
docker compose logs -f dashboard
docker compose logs -f worker
docker compose down
powershell -ExecutionPolicy Bypass -File .\worker\run-windows.local.ps1

Architecture Notes

• Control Plane: Backend API, Streamlit dashboard, PostgreSQL, Redis, n8n,

approval/documentation/upload flow.

• Worker Plane: pull-based Python worker that fetches jobs from the Backend API

and performs heavy generation/rendering only when the worker machine is on.

• Storage defaults to local filesystem and is designed to support Nextcloud

WebDAV or Google Cloud Storage later.

• LLM provider router can use mock, Gemini CLI, Claude Code, and Codex CLI

fallback depending on environment configuration.

• AI quality packs under ai_quality/ provide task-specific rules for topic,

script, metadata, fact check, image, and revisions.

• Rendering and media pipeline references FFmpeg, Remotion, TTS providers, image

quality checks, and scene-level fallback behavior.

Known Pitfalls

• README contains example environment variables and local paths. Do not copy

credentials or user-specific secrets into public/human-facing outputs.

• Some flows depend on logged-in CLI sessions, local worker machines, external

API credentials, or Docker services.

• Automated tests were not obvious in the current scan; verification is mostly

runbook/manual QA unless tests are added later.

• Production deployment, real YouTube upload status, channel count in operation,

revenue, and public metrics are Needs confirmation.

Current Work

• Personal portfolio/onboarding material is being generated from this repo.
• TODO.md shows several image prompt, caption timing, dashboard metrics,

recommendation, and n8n auto-upload tasks completed; remaining items include SFX, real cooking/history media sources, furigana handling, title A/B testing, and final mp4 visual evaluation.

Decisions

• Treat this as a systems/automation/AI pipeline portfolio item.
• Public portfolio wording should be generalized; internal onboarding can include

more architecture detail but still must not expose secrets.

Open Questions

• Real production deployment status: Needs confirmation.
• Which channels are live versus mock/demo: Needs confirmation.
• Whether real YouTube uploads are enabled in current operation: Needs confirmation.